In the last few years, rapid accumulation of genome sequences and protein structures has been paralleled by major advances in sequence database search methods. The powerful Position-Specific Iterating BLAST (PSI-BLAST) method developed at the NCBI formed the basis of our work on protein motif analysis. In addition, Hidden Markov Models (HMM) and protein structure comparison methods were applied. During last year, we made further progress in detailed analysis of the classification, evolution, and functions of several classes of proteins. In particular, we studied in detail the protein domains that are involved in prokaryotic toxin-antitoxin cassettes and in systems of defense against selfish elements. We proposed previously that the group of so-called Cas proteins comprise a novel DNA repair system. The association of the cas genes with CRISPR and, especially, the presence, in CRISPR units, of unique inserts homologous to phage and plasmid genes make us abandon this hypothesis. It appears most likely that CASS is a prokaryotic system of defense against phages and plasmids that functions via the RNAi mechanism. The functioning of this system seems to involve integration of fragments of foreign genes into archaeal and bacterial chromosomes yielding heritable immunity to the respective agents. However, it appears that this inheritance is extremely unstable on the evolutionary scale such that the repertoires of unique psiRNAs are completely replaced even in closely related prokaryotes, presumably, in response to rapidly changing repertoires of dominant phages and plasmids. We also used computational methods to identify a novel toxin-antitoxin system that is predicted to function via RNA binding or cleavage.